US lawmakers demand human control to prevent AI-launched nukes

Lawmakers in the House of Representatives have proposed legislation to ensure human control over America’s nuclear arsenal in response to concerns about the potential for AI systems to launch autonomous nuclear attacks. Rep. Ted Lieu, along with bipartisan support, has introduced an amendment to the 2024 defence policy bill that mandates a system be put in place by the Pentagon to require ‘meaningful human control’ for the launch of any nuclear weapon. The amendment defines human control as the final decision-making authority in selecting and engaging targets, including the timing, location, and method of attack.

Senior military leaders have expressed their commitment, emphasising that AI is intended to assist in data assessment and provide options for human decision-makers rather than taking autonomous actions. However, the growing bipartisan support for Lieu’s amendment indicates lawmakers’ increasing concerns regarding the potential for AI to make decisions rapidly without sufficient human oversight. The amendment, part of the National Defense Authorization Act (NDAA), has garnered support from both Republican and Democratic lawmakers.

Various other amendments related to AI have been proposed. Rep. Stephen Lynch has proposed an amendment requiring the Defense Department to adhere to the Biden administration’s guidance on responsible military use of AI. Additionally, there are proposals for collaboration on AI research between the United States and Israel, as well as establishing a process for testing and evaluating large language models for accuracy, bias, and disinformation.